1. Field of the Invention
The present invention relates to encoding/decoding a digital signal, and, more particularly, to a method of and an apparatus of encoding/decoding a digital signal using linear quantization by sections.
2. Description of the Related Art
A waveform including information is an analog signal in which amplitude of the waveform changes continuously over time. Therefore, an analog-to-digital (A/D) conversion is needed in order to express the waveform as a discrete signal. Two processes are required to perform the A/D conversion. The first is a sampling process in which the amplitude of the analog signal is sampled, and the other is an amplitude quantizing process in which the sampled amplitudes are replaced with the nearest value that is used by a device in reproducing a digital signal. That is, in the amplitude quantizing process, an input amplitude x(n) is converted into y(n), which is an element included in a finite collection of amplitudes, in time n.
When storing/restoring audio signals, according to a recent development in digital signal processing technology, a conventional audio signal is converted into a pulse code modulation (PCM) data signal, which is a digital signal, after a sampling and quantizing operation, and is stored in a recording/storing medium such as a compact disc (CD) or a digital audio tape (DAT). Then, the stored signal is reproduced and listened to again according to the needs of a user. Such storing/restoring of audio signals is widely known and used by the general public. The storing/restoring method using the PCM data improves sound quality and overcomes the problem of deterioration, which occurs according to the storage period, compared to an analog method used in for example, long-play record (LP) or a tape. However, the large size of digital data subsequently has brought about problems of storage and transmission.
To solve such problems, methods such as differential pulse code modulation (DPCM) and adaptive differential pulse code modulation (ADPCM) have been developed to condense digital audio signals. There have been efforts to decrease the amount of data in digital audio signals using such methods, but there are large variations in the efficiency of the digital audio signals depending on the types of the signals. Recently, a method of decreasing data using a psychoacoustic model of humans is being used in a moving pictures experts group (MPEG)/audio technique standardized by the International Standard Organization (ISO) and an alternating current (AC)-2/AC-3 technique developed by Dolby. These methods play a big role in efficiently decreasing the amount of data while maintaining the characteristics of signals.
In a conventional audio signal condensing technique, for example, MPEG-1/audio, MPEG-2/audio, or AC-2/AC-3, signals in the time domain are grouped into blocks of a predetermined size and converted into signals in the frequency domain. Then, scalar quantization is performed on the converted signals using the psychoacoustic model. The scalar quantization technique is simple, but scalar quantization is not the most suitable choice even if an input sample is statistically independent. Of course, scalar quantization is even more unsuitable if an input sample is statistically dependent. Therefore, no-loss encoding (e.g. entropy encoding) or encoding including some type of quantization adjustment is performed. Consequently, the condensing technique is quite complicated compared to the method of storing simple PCM data. Also, a configured bit stream includes side information to condense signals in addition to quantized PCM data.
The MPEG/audio standard or the AC-2/AC-3 method provides virtually the same sound quality as a CD with a bit ratio of 64-384 Kbps, which is ⅙ to ⅛ less than a bit radio used in the conventional digital encoding method. As such, the MPEG/audio standard is predicted to be a standard that will play an important role in storing and transmission of audio signals in, for example, digital audio broadcasting (DAB), Internet phones, audio on demand (AOD), and multimedia systems.
In the MPEG-1/2 audio encoding technology, after performing a subband filtering operation, a subband sample is linearly quantized using bit allocated information that is suggested in the psychoacoustic model, and completes the encoding using a bit packing process. In the quantizing process, a linear quantizing device provides an optimum efficiency when distribution of data is uniform. However, the actual distribution of data is not uniform, but is closer to a Gaussian or Laplacian distribution. In this case, a quantizing device is designed to fit each distribution, and an optimum result may be achieved by minimizing in a mean squared error (MSE).
A general audio encoder such as an advanced audio coder (AAC) of MPEG-2/4 uses a nonlinear quantizing device of X4/3. The AAC is designed in consideration of a sample distribution of a modified discrete cosine transform (MDCT) and the psychoacoustic perspective. However, the encoder is highly complex due to the characteristics of a nonlinear quantizing device. Therefore, the AAC generally cannot be used as an audio encoder that requires low complexity.